Subnautica did not succeed despite being built on Unity. It succeeded because Unity forced it to be scary in a cheap, legible way.

Unknown Worlds Entertainment built Subnautica (full release January 23, 2018 for Windows and macOS) and its spin-off Below Zero (May 14, 2021) on Unity, then moved the sequel Subnautica 2 (early access May 14, 2026 for Windows and Xbox Series X/S) to Unreal Engine 5. That switch is the honest part of the story: Unity was the right boring tool for a single-player oxygen nightmare, and the wrong tool for 4-player co-op open water. Subnautica infobox lists Engine: Unity, Below Zero infobox lists Engine: Unity, Subnautica 2 infobox lists Engine: Unreal Engine 5.

The series has now passed over 15 million copies sold worldwide. The first game alone had sold over 5 million copies by January 2020. Search demand explains why this engine question matters: Google and Bing both autocomplete "was Subnautica made in Unity," "what engine is Subnautica 2 made in," and "is Subnautica a Unity game" with 10+ variants checked November 12, 2026. Everyone assumes an alien ocean needs custom tech. It used phone-game tech instead.

Why a Half-Life mod team picked Unity in 2013

Unknown Worlds came from Natural Selection, a Half-Life mod, then Natural Selection 2 on its own Spark/Evolution engine. Founder Charlie Cleveland said he was exhausted after Natural Selection 2 and inspired by Minecraft's rejection of traditional progression. The team announced Subnautica on December 17, 2013 and entered Steam early access in December 2014 for Windows, June 2015 for macOS, May 2016 for Xbox One.

My take — and this is where I split from "Unity is just cheap" explainers: Unity was not only cheap, it was a distribution weapon. Unity's pitch since 2005 was democratized, cross-platform development with a C++ runtime and C# scripting API, popular for indie and mobile. For an underwater survival game that lives or dies on word-of-mouth, running on Intel HD 4600, 4 GB RAM minimum, 20 GB storage mattered more than photoreal caustics. That floor let Subnautica reach school laptops, then PS4/Xbox One in December 2018, Switch/PS5/Xbox Series X/S in May 2021, iOS/Android on July 8, 2025. Horror that needs a $2,000 PC is a tech demo. Horror that runs on a toaster is a meme, a Let's Play, a wiki.

The other boring advantage: single-player prefabs. Subnautica is strictly single-player with four modes — Survival, Freedom, Hardcore permadeath, and Creative. No replication, no server auth, no anti-cheat. A 10-person distributed team scattered from the US to Thailand and Australia could iterate on creatures, bases, and story in Unity's editor without building netcode. Below Zero reused the same deal.

What Unity actually bought: oxygen as a loading screen

Most retrospectives list "easy prototyping." That is true and useless. Here is what Unity concretely enabled:

1. Oxygen turned streaming limits into game design. Subnautica's oceans run from sunlit reefs to deep trenches, lava fields, and bioluminescent rivers, plus two islands and a day-night cycle. Unity in 2014-2018 was not great at seamless large-world streaming. Unknown Worlds made that a virtue: your oxygen tank forces you back every 30-135 seconds early on. You never notice pop-in because you are counting breaths. Bases, the Seamoth, and later the Cyclops are just bigger oxygen bottles. If you want to feel that gather-return-upgrade loop in a browser without the 20 GB install, Adventure Miner distills it well — mine until the backpack is full, sell at the dock, buy capacity, push to the next island. Subnautica is that loop with teeth.

2. Hand-built dread instead of procedural soup. Cleveland cited Minecraft, but Subnautica rejected infinite procedural terrain for a hand-placed 4546B. Every wreck, Degasi base, and Reaper patrol is authored. That choice fits Unity's prefab workflow and creates wiki-able knowledge — "Reaper behind the Aurora" — which is free marketing. Procedural would have been cheaper to build and more expensive to scare with.

3. Hull integrity taught pressure without a physics lecture. Build bases on the sea floor and manage hull-integrity as depth and pressure increase. It is fake engineering — reinforcements and bulkheads as hit points — but it reads instantly. Beaver Builder is a goofy but honest trainer for the same instinct: gather, build against flooding, watch the structure fail if you overextend. Subnautica players learn the same lesson with moonpools instead of dams.

4. Vehicles as moving bases solved the home problem. The Seamoth, PRAWN Suit with drills, torpedoes, propulsion cannons and grapples, and the Cyclops submarine let you carry oxygen, storage, and safety with you. Technically, this kept the player inside a small bubble of loaded content while the ocean streamed around them. Design-wise, it turned "go deeper" into a tech tree: fins → tank → Seamoth → Cyclops → PRAWN → rocket.

5. Darkness + audio did what shaders could not. Unity could not out-render CryEngine water in 2018. So Unknown Worlds made night genuinely dangerous — predators come out, safe shallows turn treacherous — and let the Reaper's roar do the work of ray tracing. Big Shark trains the same reading skill in two minutes: eat what is smaller, avoid what gurgles louder, grow only when you respect the food chain. Leviathans are Big Shark with reverb and trauma.

What it cost: the Unity tax every veteran recognizes

Single-player Unity had real bills:

  • Performance and save bloat. Large bases, parked vehicles, and hundreds of containers bloated saves and hitch on base load. The minimum spec is honest — Intel HD 4600 will launch the game — but "launch" and "enjoy the Lost River at 60 fps" are different promises.
  • No multiplayer foundation. The engine choice baked in single-player assumptions: pause-on-inventory, save-anywhere, time-freeze crafting. Adding co-op later would mean rewriting oxygen, story triggers, and the Quarantine Enforcement Platform logic for 2-4 timelines.
  • VR as a checkbox. VR support for HTC Vive and Oculus Rift with keyboard/gamepad exists, but never got the UI love it needed. Unity made VR easy to enable and hard to make comfortable underwater.
  • Horror dilution through familiarity. Once you know Reaper spawns, fear becomes route planning. Hand-built maps are spoilerable in a way procedural horror is not.

Unknown Worlds paid those taxes for seven years because the alternative — custom engine like Spark, or Unreal 4 with a larger team — would have delayed early access and killed the feedback loop that made Subnautica good.

Why Subnautica 2 abandons Unity for Unreal Engine 5

Subnautica 2 supports single-player and 4-player co-op, the first multiplayer in the series, on a new planet Proteus. It adds DNA-modifying biomods and a reprint/respawn fiction to justify co-op death. It is made with Unreal Engine 5.

That is not a fashion upgrade. Three pressures force it:

  1. Replication. Unity can do multiplayer, but Subnautica's save-everything, build-anywhere world needs robust replication, interest management, and host migration for subs that move. Unreal's replication framework plus a larger Krafton-era team after the October 2021 Krafton acquisition is a saner base than retrofitting Unity single-player saves.
  2. Density without draw-call hell. Coral reefs, the World Tree, and 4-player structures need Nanite-style geometry and Lumen-style global illumination more than Subnautica 1's flat-shaded stylization did. UE5's World Partition also fits an ocean that is vertical, not flat.
  3. Hiring and outsourcing. After Krafton, Unknown Worlds hires in a market where UE5 experience is easier to scale than bespoke Unity ocean tooling.

My contrarian take: this upgrade risks the "ugliness moat." Subnautica 1's jank — inventory tetris, Cyclops getting stuck on mushrooms, that 3-second freeze when you surface too fast — made the ocean feel indifferent. Polished UE5 co-op with pings, shared bases, and revives could make Proteus feel like a cruise. The original was camping alone with a leaking tent. If Unknown Worlds smooths every edge to satisfy live-service content-volume reviews — the May 2025 Krafton memo asking for 30% more content and the July 2025 leadership fight and delay to 2026 show that pressure is real — they may gain players and lose dread. Early access launching May 14, 2026 after lawsuits, a March 2026 court order, and leaked builds will test whether co-op horror can stay lonely.

What indie devs should actually steal

  1. Make your engine limit your UI. Oxygen is health, timer, tutorial, and difficulty slider in one number. If your engine streams poorly, make the player want to come back often.
  2. Ship on toasters, scare on YouTube. Low specs are not charity; they are distribution. Subnautica's Switch and mobile ports (2021, 2025-2026) extended a 2018 game's tail past most AAA.
  3. Author fear, don't generate it. Hand-place 10 perfect scares instead of 1,000 random ones. Wikis will document them and sell your game for you.
  4. Vehicles are level design. If traversal is boring, give the player a louder, hungrier traversal that needs fuel and parking.
  5. Plan the sequel engine early. If you ever want co-op, do not build single-player pause-and-save into every system. Isolate story state now or you will rewrite everything later — which is exactly what Unknown Worlds is doing by leaving Unity.

If you are prototyping an underwater loop this weekend, play the trainers in order: Adventure Miner for the ore-to-upgrade pacing, Ages: Build & Craft for village-to-city base progression from primitive to modern, and Beaver Builder for "build against water or drown." Then load Subnautica and notice how familiar the first 30 minutes feel — Life Pod, scan limestone, build O2 tank, fear the dark.

Unity did not make Subnautica beautiful. It made it breathable, buildable, and cheap enough to drown in.